Step 2B: Advanced Methods for Projecting a Caseload

Other Ways to Estimate a Projected Caseload

There are several accepted ways of projecting future caseloads from historical caseloads. Step 2 presented the simplest method: using a recent year’s caseload as your Projected Caseload. More advanced projection methods attempt to both minimize the influence of past years’ caseload anomalies and account for trends that might impact the next year’s caseload. Popular advanced projection methods include: a multi-year average, an adjusted multi-year average, and a statistical model.



A. Multi-year Average

A multi-year average is one projection strategy to address caseload anomalies. Using a multi-year average, instead of only the most recent year’s caseload data, reduces the influence of anomalies or one-time events, like an unusually high or low number of murder cases in a year. 

The table below shows five years of historical caseload data for Fictional Jurisdiction (2020-2024). 

 

The 2020 and 2021 caseloads show the impact of the COVID-19 pandemic: caseloads rebounded in 2022, stabilized in 2023, and then grew in 2024. 

To use a multi-year average as the projection for next year, you simply average several past years caseloads for each case type. In the example below, Fictional Jurisdiction used a three-year average to project its 2026 caseload, using its 2022, 2023, and 2024 caseloads by case type.  

 

 

What is a Moving Multi-Year Average?

When a jurisdiction recalculates its multi-year average year after year it is called a moving (or rolling) average. Each year, the jurisdiction adds the most recent year’s data to the average and drops the oldest year’s data from the average.   

In our above example, Fictional Jurisdiction used a three-year average to generate a projected 2026 caseload. Fictional Jurisdiction used its 2022, 2023, and 2024 to calculate its 2026 caseload.  If, the following year, Fictional Jurisdiction used data from 2023, 2024 and 2025 to project its 2027 caseload (dropping the 2022 data and adding in 2025 data), Fictional Jurisdiction would be using a moving three-year average. Many state court systems use a three-year or five-year moving average to project their caseloads and calculate their staffing needs.  

B. Adjusted Multi-Year Average

In some instances, you may have reliable information about legal or practice changes that have either impacted past caseloads or are likely to impact the next year’s caseload. In that situation, you might begin with the multi-year average (or the previous year’s caseload) and then adjust your caseload estimates to reflect those changes. 

 

For example, Fictional Jurisdiction adopted two recent law changes that are likely to impact caseloads. To account for these changes in its Projected Caseload, Fictional Jurisdiction begins with a three-year average (shown above) but then makes some adjustments to the resulting average to account for these law changes. 

Adjustment 1: In 2023, Fictional Jurisdiction adopted a new sentencing scheme that recategorized some misdemeanors as felonies. The provider believes that this sentencing change accounts for the increase in 2024 Felony-Low cases and the decrease in 2024 Misdemeanor-High cases. Rather than using the three-year average for all case types, the provider handles the impacted cases differently, using the 2024 Felony-Low (11,914 cases) and Misdemeanor-High (4,208 cases) caseloads to project their 2026 caseload.

 

{New image coming}

 

Adjustment 2: Fictional Jurisdiction recently announced a limited-scope second-chance misdemeanor program for people who have been charged with petty offenses and have no prior criminal history. The court clerk automatically identifies eligible arrested people and defers their cases. Unless they are rearrested within six months of deferral, these people never appear in court. Based on past caseload data, the provider estimates that this will reduce their Misdemeanor-Low caseload by 20%. The provider therefore estimates the projected 2026 Misdemeanor-Low caseload at 80% of the three-year average Misdemeanor-Low caseload or 9,598 cases. 

 

[New Image coming}

The resulting adjusted three-year average caseload for Fictional Jurisdiction is shown below.

 

C. Statistical Models

Statistical models can use trends and patterns in previous caseloads to predict future caseloads. They can also identify and exclude or address anomalies. Some models can even take account other data, such as crime statistics or arrest rates. Common statistical models for caseloads include Autoregressive Integrated Moving Average (ARIMA) and Holt-Winters forecasting. While statistical models may be more accurate than a multi-year average, they generally require significantly more data. Note: You should only consider a statistical model if you have an experienced researcher to do these calculations for you.